The same reasoning applies to visual processing.
Object recognition has always been one of our strongest points. Hand us a photo of an object, say a bicycle or a fire hydrant, then ask us to find it in another photo and we’ll have no problems. This simple task is actually incredibly difficult for a computer to pull off, since the target object may be rotated differently, obscured behind another one, differently coloured, or even have a texture overlaid on it. Tomaso Poggio of MIT thinks it’s as difficult a feat as simulating intelligence in general.
He should know: he’s partially cracked that problem.
In 2007, he released two papers that detailed a new method of object recognition. For one of those papers, he did exactly the task I outlined above. The new algorithm was able to track down an object about 97% of the time, though it could range from 93% to 99.8% accuracy depending on the exact task. Unlike most other algorithms, which specialize in finding only one type of object, his method works equally well on a wide variety of objects.
There’s good reason for that: it’s modelled on the human visual system. His second paper demonstrates just how closely that model follows reality, by pitting man against machine. Poggio sat human beings down in front of a computer, flashed them a single image for 1/50th of a second, then asked the humans if there was an animal in the scene or not. We’re literally born for this task; how fast you can tell if that’s a wild animal or a branch dropping towards you determines how likely you are to survive and pop out offspring. Unsurprisingly, human beings do pretty well at this, getting it right around 80% of the time.
Poggio’s algorithm was then used to classify each image. The results? It guessed correctly 80% of the time. Eerily, when the researchers analysed the images the algorithm did poorly on, they discovered that the humans had struggled on those images as well.
This algorithm can still be improved. Our brains use feedback from other parts of the brain to improve our guesses further, something this method doesn’t handle. It also took much longer for this algorithm to come to a conclusion than our brains did, but that’s only a temporary problem. Computers improve in speed much faster than human brains do, and more efficient programming should reduce the number of necessary calculations.
Still, Poggio’s work hints very strongly that there’s no magic to our visual system.
Sounds may be another matter, though. Not just any sound, though, but the semi-repetitive collections of sound we call music. Humans have spent centuries, probably even millennia, creating harmonies and melodies for no other reason than pleasure. No other species can dare make that claim.
Well, except birds. And maybe whales. Oh, and gibbons.
But before we get into the details, we first have to settle what music is. The suboscine branch of the bird family can have elaborate calls, but those are hard-wired into their genes. You can separate them from their parents, play the songs of other birds as often as you want, and they’ll still chirp out their innate tune. Most people would not consider this music; there must be an element of creativity involved, and while genes can produce variation through mutations, that happens on too long a time scale to qualify.
Songbirds are a different feather. Play them a tune at the right age, and they’ll pick it up and use it as their own. Deprive them of music, and they’ll sing poorly or not at all. While better, this still doesn’t quite qualify as a creative act since they’re just copying the songs they heard. Changes will happen over time due to accident, faster than they would through genes, but still not fast enough.
Not all songbirds are born alike, though. The Indigo Bunting will pluck a song out of thin air, with no resemblance to anything it’s heard before, then slowly mix in fragments from nearby competitors until it becomes a variation on a theme. Mockingbirds got their name from a remarkable ability to imitate sounds in their environment, everything from the calls of insects to the ring of a cell phone, which are then incorporated into their songs. Both species can be considered creative.
Both could also be dismissed as too greedy. Birdsong is primarily used to attract mates, warn about predators, and establish territory. It also makes a handy show of fitness; sick birds have difficulty carrying a tune, and it puts them at risk of an attack by predator. The music that human beings make has much purer motives, and is rarely used to show off.
Sorry, but I couldn’t keep a straight face while writing that last line. One of the leading theories of why we make music is that it’s a show a reproductive fitness. Humans can’t sing or play an instrument very well if they’re sick, either, and we frequently use music to set a romantic mood. As Geoffrey Miller of the University of New Mexico has pointed out, musical output peaks and declines with sexual ability, and a whopping 40% of all lyrics relate to sex or romance. Musicians are usually considered sexually desirable.
Consider Jimi Hendrix, for example. This rock guitarist extraordinaire died at the age of 27 in 1970, overdosing on the drugs he used to fire his musical imagination. His music output, three studio albums and hundreds of live concerts, did him no survival favours. But he did have sexual liaisons with hundreds of groupies, maintained parallel long-term relationships with at least two women, and fathered at least three children in the U.S., Germany, and Sweden. Under ancestral conditions before birth control, he would have fathered many more. […] As Darwin realized, music’s aesthetic and emotional power, far from indicating a transcendental origin, point to a sexual-selection origin, where too much is never enough. Our ancestral hominid-Hendrixes could never say, “OK, our music’s good enough, we can stop now”, because they were competing with all the hominid-Eric-Claptons, hominid-Jerry-Garcias, and hominid-John-Lennons. The aesthetic and emotional power of music is exactly what we would expect from sexual selection’s arms race to impress minds like ours.
(“Evolution of human music through sexual selection,” by Geoffery Miller. Published in “The origins of music,” edited by N. L. Wallin et al, 2000. )
Most damning of all is the genetic evidence. If music was related to reproduction, we should expect to find genes that control it. Not only do those genes exist, but they’re almost identical to the genes that give birds the ability to create songs.
The identification of FOXP2 as the monogenetic locus of a human speech disorder exhibited by members of the family referred to as KE enables the first examination of whether molecular mechanisms for vocal learning are shared between humans and songbirds. […] In support of this idea, we find that FOXP1 and FOXP2 expression patterns in human fetal brain are strikingly similar to those in the songbird, including localization to subcortical structures that function in sensorimotor integration and the control of skilled, coordinated movement. The specific colocalization of FoxP1 and FoxP2 found in several structures in the bird and human brain predicts that mutations in FOXP1 could also be related to speech disorders.
(“Parallel FoxP1 and FoxP2 Expression in Songbird and Human Brain Predicts Functional Interaction,” by Ikuko Teramitsu et al. The Journal of Neuroscience, March 31, 2004, 24(13):3152-3163)
There’s still an objection to be made, even if we agree with Miller and others. An evolved trait may be used differently at different times. Music in human beings may have started as a show of fitness, but it need not stay that way. After all, very few people actively pursue a career in music; the majority instead write songs privately, for their own enjoyment. Human beings may have once sung for sex, but nowadays we’re more likely to sing for ourselves.
Against this stands the Brown Thrasher. Birdsong comes in roughly five flavours: mating song (“I’m here, and I’m sexy!”), companion calling (“I’m here, where are you my mate/friend?”), begging by young birds (“GIMMMIE FOOOOD NOOOOOOOW!!”), trespass threats (“Get out of my area, you upstart, or else!”), and predator alerts (“I see something dangerous!”). Sometimes the lines can blur a bit (“Everybody, come help me harass this predator!”), and some species have multiple calls within each flavour (“Head’s up, it’s a predator from the sky!”), but a grand total of a dozen or two should be more than enough for most birds. And it is, generally.
So why does the Brown Thrasher have a library of 2,000 calls? To give that a baseline, the average vocabulary of a human being consists of 10,000 words.
It’s tempting to dismiss all that variation as invention. The Thrasher may be taking a “base” song and improvising new versions of it. If this is the case, we’d expect very few songs to be repeated; instead, Thrashers can recall a song they tweeted nineteen days earlier.
Alternatively, that vast song repertoire may be a way to show off to the opposite sex. There’s a problem, however; most female songbirds are not attracted to males with a giant songbook. One study showed that Brown Thrashers were actually more interested in a limited sample of Thrasher calls than the full collection, provided they displayed more versatility in singing ability.
It’s tough to draw a definitive conclusion from a single study, so I won’t. What I will say is that it’s plausible the Brown Thrasher’s vast library is more for personal kicks than practical use.
Intrapersonal and Self-Awareness
Humans are quite good at understanding the personal. We’ve got an entire branch of science devoted to it, named psychology. Philosophers from Plato onward have valued looking inward, to discover what we really are like. Surely no other species can come close to us here.
So far as we know, none has. It’s hardly their fault, though; we have only two ways to learn about the inner lives of others, by direct communication and indirect brain scanning. With no way to ask other animals how they feel, and quite different brain structures between us, plumbing the depths of other species’ cognition ranges from difficult to impossible.
It doesn’t help that we tend to project ourselves onto other creatures. Alexandra Horowitz conducted a study that asked dog owners to forbid their pets to eat a treat. When the humans left the room, Horowitz randomly fed some of the dogs that forbidden fruit; when they came back, she randomly told some of the owners that the dog had eaten the treat. The humans that were told their pet had broken their order thought their dogs looked guilty, even if they never ate the treat. When punished, the pets that looked most guilty were actually the ones which never got a lick at the prize. Any interpretation of animal behaviour has to be done very, very careful to filter out our inner biases.
We do, however, have a proxy for inner knowledge: knowledge of the self. If you have no concept of “you,” there’s no self to learn about. And once you realize there’s a “you” there, curiosity will drive you to give it a quick once-over, at minimum.
The standard test for self-awareness is pretty simple: place an animal in front of a mirror, and let them get used to it. Then put them to sleep, paint a dot on an area of their body that they can’t normally see, then wake them up and place them by the mirror. If they try to rub off or touch the dot, they must know the animal in the mirror is actually themselves, and must be capable of mapping between the image and themselves. Human children pass this test easily, as do all primates, elephants, dolphins, and European magpies. Other species, such as pigs and pigeons, fail this test but can demonstrate that they know the image in the mirror reflects reality. In the case of pigeons, this can even be used to “train” them for the test, resulting in a pass.
Those last results have been used to criticise the test; perhaps a species just doesn’t care about cleaning off the dot, leading researchers to falsely conclude it isn’t self-aware. Another problem is that self-recognition may not be tied to self-awareness; humans with prosopagnosia cannot recognize themselves, yet clearly are self-aware. Note however that both arguments lead us to conclude there are more self-aware species than we realize, not less.
 “Robust Object Recognition with Cortex-Like Mechanisms ,” Thomas Serre et al. IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol 29 No 3, March 2007 .
”A feedforward architecture accounts for rapid categorization,” Thomas Serre et al. Proceedings of the National Academy of Sciences, vol. 104 no. 15 6424-6429, April 10, 2007.
Boughey, M. J. and Thompson, N. S. (1981), “Song Variety in the Brown Thrasher (Toxostoma rufum). Zeitschrift für Tierpsychologie, 56: 47–58. doi: 10.1111/j.1439-0310.1981.tb01283.x